🚀 go-pugleaf

btonkin@lindblat.cc.monash.edu.au (Bruce Tonkin) writes:

>Does anyone else use Wang DAT tape drives (4mm tape cartridge) with
>SUN workstations for backup?

We have three or four Suns and about 10 GB of disk and back up onto a
WangDAT 2600 drive connected to a SPARCserver II.  You need a kernel patch
(see below).  I did some tests using our data (maybe 20% of which is
compressed) and found you can store about 4 GB of data on a tape in compressed
mode.  Your milage will vary depending on your type of data.

It takes about an hour to dump 1 GB and uses about 20% of the CPU according
to top so we can run dumps during the day if need be.

You need to experiment with dump to work out good parameters.  Just set the
tape length to be something really large and/or set the tape density.
I designed the dumps no greater than about 3 GB so we should never reach
the end of a tape.

We put multiple dumps on each tape using 'mt' to move between the files.
If you like I'll send you the scripts, (they're nothing fancy though).

Overall we are quite happy with the WangDAT.  I'd like to test the integrity
of our dump tapes on another WangDAT drive.  Would you like to arrange to swap
test tapes and see if we can read each others tapes ?

Unfortunately Wang use a prop. compression scheme and I don't know anybody
else with a WnagDAT drive.  I believe Sony now make a DAT drive using
a standard compression that many manufacturers are supporting.

Anthony Botham
SysAdmin
CITRI / Computer Science, RMIT
Melbourne

ph. (03) 282-2484
asb@citri.edu.au

---------------------------------------------------------------------------
---------------- Begin Included Message -----------------

From: metcalf@catfish.LCS.MIT.EDU (Chris Metcalf)
Newsgroups: alt.sys.sun
Subject: RESULT: attaching 4mm DAT to Sparcstation
Date: 16 Oct 91 13:01:03 GMT
Lines: 53
Originator: metcalf@catfish.LCS.MIT.EDU

I wrote:
> We have been happily using an Exabyte for backups for a while now, and
> recently decided to add a WangDAT Model 1300 as well.  [...]
> Unfortunately, the throughput is absymal [on our Sparcstations].

Thankfully, the answer turns out to be fairly simple.  Thanks to:

	wolfgang@wsrcc.com (Wolfgang S. Rupprecht)
	jpd@pc.usl.edu (Dugal James P.)
	arul%saturn@sdsu.edu (Arul Ananthanarayanan)

The problem is that the kernel IDs the tape drive by using the vendor
ID field returned by the SCSI protocol, and if it doesn't match the ID 
to anything in the st_drivetypes[] array it defaults to a non-buffered,
very slow default drive type.  To correct the problem, just add an
entry for your particular drive in (sun4c): /usr/sys/scsi/targets/st_conf.c:

*** 1.1	1991/10/11 20:18:40
--- st_conf.c	1991/10/15 15:43:47
***************
*** 178,184 ****
  	500, 500,
  	{ 0x01, 0x02, 0x06, 0x06},
  	{ 0, 0, 0, 0 }
  }
  };
  int st_ndrivetypes = (sizeof (st_drivetypes)/sizeof (st_drivetypes[0]));
  
--- 178,192 ----
  	500, 500,
  	{ 0x01, 0x02, 0x06, 0x06},
  	{ 0, 0, 0, 0 }
+ },
+ /* WangDAT Model 1300 3.81mm helical scan cartridge */
+ {
+ 	"WangDAT 3.81mm Helical Scan", 7, "WangDAT ", ST_TYPE_EXABYTE, 512, 
+ 	(ST_BSF | ST_BSR | ST_LONG_ERASE | ST_AUTODEN_OVERRIDE),
+ 	5000, 5000,
+ 	{ 0x00, 0x00, 0x00, 0x00, },
+ 	{ 0, 0, 0, 0 }
  }
  };
  int st_ndrivetypes = (sizeof (st_drivetypes)/sizeof (st_drivetypes[0]));

With the patch in place, the drive dumps out at about 160 KB/sec
(almost 10 MB/min).  Note that for my WangDAT entry, "mt status" will
still report it as an Exabyte EXB-8200 8mm tape drive since I set the
ST_TYPE_EXABYTE flag in the new structure.  This may or may not actually
be necessary or helpful, but it does seem to be working.

-- 
			Chris Metcalf, MIT Laboratory for Computer Science
			metcalf@masala.lcs.mit.edu // (617) 253-7766

---------------- End Included Message -----------------

-- 
Paul Ziemba  paul@alantec.com	alantec!paul  408.955.9000

Current Nemesis: *.ps